The strengths of our analysis of 3154 middle-aged U.S. women include the large sample size, the multiethnic composition, and the ability to analyze the women with history of menstrual regularity. SWAN participants were required to have a menstrual period within the three months prior to the study entry, thus effectively eliminating subjects with long-term amenorrhea. Therefore, our data point to influences on fertility that are weight-related, yet are not necessarily linked with anovulation. Another strength of our report is the multi-ethnic composition of the study population with significant inclusion of minorities that are often under-represented in clinical research.
The main finding of this cross-sectional study is the association of BMI ≥30 kg/m2 in adolescence with increased likelihood of lifelong childlessness as compared with women with normal adolescent BMI. Adolescent obesity (reported HSBMI ≥30kg/m2) remained independently associated with lifetime nulliparity and nulligravidity after adjustment for adult BMI at study entry, history of non-gestational amenorrhea, marital status, ethnicity, study site, education, and socioeconomic status. The findings were not affected by tubal or male factor infertility, use of fertility treatments, decisions to remain voluntarily childless, or preferences for same sex sexual relationship. We did not see an association between adolescent obesity and spontaneous miscarriage, which is in contradistinction to data generated in several studies and a recent meta-analysis. A potential explanation is that SWAN by design excluded women who did not have a menstrual period in the three months preceding the study entry. Thus, women with polycystic ovary syndrome were likely under-represented in SWAN as compared to general population and this may contribute to the observed lack of association between obesity and spontaneous miscarriage.
As we were unable to verify the etiology of infertility, we used nulliparity as the primary outcome as it is least likely to be affected by recall. However, there was no difference in self-report of infertility by the HSBMI category. The study questionnaire did not distinguish between primary or secondary infertility (i.e., difficulties in conceiving after a prior conception). It is possible that the adverse effect of elevated BMI on fertility varies with age. In support of this notion, a recent report by Sneed et al of 1273 women aged 22 to 44 years examined the influence of age on IVF success rates (17
). In this study, BMI was observed to be an effect modifier of the relationship between age and conception. A higher BMI was a negative predictor of pregnancy only in the younger age group, diminishing considerably in the late 30's. It is possible that our inability to assess the temporal sequence between the history of infertility and body mass index contributed to the finding of no difference in reporting of infertility by HSBMI category.
Limitations of the study include the cross-sectional nature of the data that, in general, should be viewed as hypothesis-generating and descriptive rather than establishing a cause-and-effect relationship. In particular, certain variables are subject to prevalence/incidence bias, because we could not be certain that a given factor precedes the outcome. However, our exposure (reported HSBMI) and outcome (lifetime parity and gravidity) variables have been chosen to represent events that are unlikely to have changed recently for most participants. In addition, it should be pointed out while smoking did not appear to have an influence on the outcomes of interest, this effect is likely only pertinent to this study and not the general population.
Another limitation is the fact that determination of high school weight was made by self-report and recall. While there was not a validation study in SWAN, the literature suggests that agreement between self-report and measured weight is reasonable, with one study reporting a correlation coefficient of 0.84 between recalled and measured high school BMI (18
). Data from the Newton Girls Study provided similar results with a correlation of 0.61 (p<0.001) between the BMI percentile at menarche and recalled body size at menarche (19
). We used the WHO categories for adult BMI cut-points in our assessments. However, CDC recommends using the percentile BMI for age and gender to define overweight status in childhood (4
). Study participants were asked to recall their weight when they left high school and hence were approximately 17-18 years of age. In adolescence, the 85th
percentile roughly approximates BMI of 25 kg/m2
and thus adult cutoffs are appropriate (20
). Although we have evaluated several social factors that may influence childbearing, we did not have an ability to consider other possible societal issues (i.e. discrimination and prejudices against obese individuals or simply fewer sexual partners) that might account for the observed relation. Lastly, the distribution of BMI varies by ethnicity and is different in Asian women relative to other groups (21
). The use of the same cut-points for defining obesity regardless of ethnicity might result in misclassification of some women and, consequently, underestimate the impact of obesity.
Our study suggests a detrimental impact of adolescent adiposity on parity later in life. In 1952, Rogers and Mitchell observed a higher incidence of obesity in a group of young women ages 16-29 with amenorrhea (22
). In a subcohort analysis from the Nurses' Health Study (NHS), women with a history of higher adolescent BMI had an increased risk of anovulatory infertility over 17 years of follow-up (13
). In NHS, the relative risks for all categories of BMI above 23.9 were statistically significant, suggesting that even moderate adiposity played a role in disturbing menstrual function. Similarly, a 1958 British birth cohort study reported that obesity at age 7, defined as approximately 98th percentile based on a population life table, conferred a 78% higher risk of menstrual irregularity at age 33 (12
). While the data are overwhelming that obesity influences fertility, it should be noted the precise mechanism remains to be elucidated. Bolumar et al reported a harmful effect of obesity on reproduction only in smokers (9
); whereas Sneed et al found a detrimental effect of obesity on IVF pregnancy rates only in young patients and not in women over 35 (17
). On the contrary, Gesink et al (7
) and Jensen et al (8
) found that neither age nor smoking made an impact on the association of obesity with decreased fertility as manifested by monthly pregnancy rates. However, the uncertain mechanism notwithstanding, we herein report an independent effect of adolescent obesity on lifetime nulliparity apart from other covariates.
There is a considerable body of evidence to support the biological plausibility of the negative impact of adiposity on fertility, mainly due to attenuation of the central hypothalamic drive in overweight and obese women (23
). In a detailed evaluation of daily hormone patterns from a subcohort of the same sample as the current study, a progressive decrease in urinary luteinizing hormone, follicle-stimulating hormone, and luteal pregnanediol glucuronide was observed with increasing BMI (25
). Selective impairment in LH pulse amplitude but not frequency has been observed in obese ovulatory women (26
). An alternate theory was suggested based on peripheral impairment of the hypothalamo-pituitary-ovarian (HPO) axis, possibly via impaired endometrial receptivity (27
). However, recent data from the donor oocyte model did not observe a decrease in implantation in the obese donor egg recipients (28
), supporting the hypothesis of the hypothesis that the adverse effects of obesity on fertility are exerted at the hypothalamic-pituitary-ovarian axis levels of the reproductive system and not through an effect on uterine receptivity.
In summary, we found that self-reported adolescent overweight status is independently associated with reduced lifetime parity and gravidity in a multi-ethnic cross-sectional sample of middle-aged U.S. women. The magnitude of the effect was higher with the increasing reported HSBMI. The cross-sectional nature of our study implies that it should be regarded as hypothesis-generating: does adolescent obesity result in diminished fertility? Prospective studies are needed to corroborate our results.